People detection in low-resolution video with non-stationary background

In this paper, we present a framework for robust people detection in low resolution image sequences of highly cluttered dynamic scenes with non-stationary background. Our model utilizes appearance features together with short- and long-term motion information. In particular, we boost Integral Gradient Orientation histograms of appearance and short-term motion. Outputs from the detector are maintained by a tracker to correct any misdetections. A Bayesian model is then deployed to further fuse long-term motion information based on correlation. Experiments show that our model is more robust with better detection rate compared to the model of Viola et al. [Michael J. Jones Paul Viola, Daniel Snow, Detecting pedestrians using patterns of motion and appearance, International Journal of Computer Vision 63(2) (2005) 153–161].